Table 1.
Network architecture | ANN error (%) | SNN error (%) | Error increment (%) |
---|---|---|---|
4-layered networks (Cao et al., 2015) (Input cropped to 24 x 24) |
20.88 | 22.57 | 1.69 |
3-layered networks (Esser et al., 2016) | − | 10.68 | − |
8-layered networks (Hunsberger and Eliasmith, 2016) (Input cropped to 24 x 24) |
16.28 | 16.46 | 0.18 |
6-layered networks (Rueckauer et al., 2017) | 8.09 | 9.15 | 1.06 |
VGG-16 (ANN model based conversion) |
8.3 | 8.54 | 0.24 |
VGG-16 (SPIKE-NORM) |
8.3 | 8.45 | 0.15 |